Analysis of the Structure and Stability of a ... - ACS Publications

template, 1, is substantially destabilized relative to the substrate. * Voice mail and fax: (312) 702-7063. E-mail: [email protected]. (1) (a) Zamec...
0 downloads 0 Views 324KB Size
J. Am. Chem. Soc. 1998, 120, 3019-3031

3019

Analysis of the Structure and Stability of a Backbone-Modified Oligonucleotide: Implications for Avoiding Product Inhibition in Catalytic Template-Directed Synthesis Peizhi Luo, John C. Leitzel, Zheng-Yun J. Zhan, and David G. Lynn* Contribution from the Searle Chemistry Laboratory, 5735 Ellis AVenue, Department of Chemistry, The UniVersity of Chicago, Chicago, Illinois 60637 ReceiVed August 14, 1997

Abstract: The structural and thermodynamic origins of the destabilization of a backbone-modified DNA duplex 1, formed between d(CpGpTNTpGpC), containing a single aminoethyl group (-CH2-CH2-NH2+-) in place of the phosphodiester (-O-PO2--O-) linkage of the central TT dimer, and d(GpCpApApCpG) are investigated. Analyses for the corresponding native duplex and two other related structural analogues of duplex 1 have been compared. Duplex 1 shows a cooperative thermal melting transition that is consistent with a twostate process. At a 2 mM concentration, the melting temperature of duplex 1 is reduced by 17 °C from the native duplex, and this decrease in stability is further assigned to an unfavorable decrease in enthalpy of 7 kcal mol-1 and a favorable increase in entropy of 15 eu mol-1. NMR structural analysis shows that the modified duplex 1 still adopts a canonical B-DNA conformation with Watson-Crick base pairing preserved; however, the CH2 group that replaces the native PO2- group in the modified backbone is flexible and free to collapse onto a hydrophobic core formed by the base edges and sugar rings of the flanking TT/AA nucleosides of the duplex. This conformation is significantly different from the maximally solvent-exposed orientation of the native phosphate in DNA. The entropic origin of the 15 eu mol-1 difference between the native and the modified duplex 1 is attributed to the hydrophobic interaction between the collapsed ethylamine linkage with the hydrophobic core of duplex 1. This assignment is further supported by a favorable comparison between the observed change in entropy and the estimated value for the hydrophobic interaction around the modified region. This estimated value is based on recent experimental measurement of the hydrophobic interaction between aliphatic groups and nucleic acids as well as ethylamine solvent transfer data. The overall decrease in the stability of duplex 1 results from a decrease in base stacking and hydrogen-bonding interactions between the base pairs. A model is, therefore, proposed to explain how the change in the local backbone conformation could disrupt the long-range cooperativity of DNA duplex formation upon backbone modifications. These studies provide an approach for identifying the factors that control the stability of the nucleic acid duplex structures containing backbone modifications, with direct implication for designing antisense oligonucleotides and template-directed reactions containing non-native phosphodiester linkages.

Introduction Stephenson1

Zamecnik and first showed that synthetic oligonucleotides could be used to block gene expression by binding specifically with complementary sequences of target mRNAs. Over the past decade the efforts to modify native phosphodiester linkages to improve nuclease resistance and membrane permeation of naturally occurring oligonucleotides for therapeutic applications have been extensive.2 These neutral backbone-modified oligonucleotides generally show attenuated binding affinity to their complementary RNA strand relative to the corresponding nucleic acid.2e,f,3 These studies have therefore clearly demonstrated that it is possible to tune nucleic acid duplex stability * Voice mail and fax: (312) 702-7063. E-mail: [email protected]. (1) (a) Zamecnik, P. C.; Stephenson, M. L. Proc. Natl. Acad. Sci. U.S.A. 1978, 75, 280. (b) Stephenson, M. L.; Zamecnik, P. C. Proc. Natl. Acad. Sci. U.S.A. 1978, 75, 285. (2) (a) De Mesmaeker, A.; Waldner, A.; Lebreton, J.; Hoffmann, P.; Fritsch, V.; Wolf, R.; Miller, P. S. In Oligonucleotides: Antisense Inhibitors of Gene Expression; Cohen, J. S., ed.; CRC Press: Boca Raton, FL, 1989, pp 79-93. (b)Uhlmann, E.; Peyman, A. Chem. ReV. 1990,90, 544. (c) Carruthers, M. H. Acc. Chem. Res. 1991, 24, 278. (d) Varma, R. S. Synlett 1993, 621. (e) De Mesmaeker, A.; Altmann, K.-H.; Waldner, A.; Wendeborn, S. Curr. Opin. Struct. Biol. 1995, 5, 343. (f) De Mesmaeker, A.; Ha¨ner, R.; Martin, P.; Moser, H. E. Acc. Chem. Res. 1995, 28, 366.

with structural variation of the backbone linkage. Product inhibition severely limits template-directed reactions,4 and this insight into the control of duplex stability has recently made it possible to design reactions that catalytically translate the information encoded in DNA into different polymer backbones.5 This catalytic reaction involves the template-directed reductive ligation of an imine to a product, d(CpGpTNTpGpC), containing a single aminoethyl group (-CH2-CH2-NH2+-) replacing the normal TT phosphodiester linkage. The catalytic cycle exploited the observation that the duplex formed between the backbonemodified product and its complementary d(GpCpApApCpG) template, 1, is substantially destabilized relative to the substrate (3) Egli, M. Angew. Chem., Int. Ed. Engl. 1996, 35, 1894. (4) (a) Naylor, R.; Gilham, P. T. Biochemistry 1966, 5, 2722. (b) Orgel, L. E. Nature 1992, 358, 203. (c) Hong, J.-I.; Feng, Q.; Rotello, V.; Rebek, J. Science 1992, 255, 848. (d) Li, T.; Nicolaou, K. C. Nature 1994, 369, 218. (e) Ertem, G.; Ferris, J. P. Nature 1996, 379, 238. (f) Sievers, D.; von Kiedrowski, G. Nature 1994, 369, 221. (5) (a) Goodwin, J. T.; Lynn, D. G. J. Am. Chem. Soc. 1992, 114, 9197. (b) Goodwin, J. T.; Luo, P. Z.; Leitzel, J. C.; Lynn, D. G. InSelf-Production of Supramolecular Structures: From Synthetic Structures to Models of Minimal LiVing Systems, Fleischaker, G. R., Colonna, S., Luisi, P. L., Eds.; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1994; pp 99104. (c) Zhan, Z.-Y. J.; Lynn, D. G. J. Am. Chem. Chem. Soc. 1997, 119, 12420-12421.

S0002-7863(97)02869-2 CCC: $15.00 © 1998 American Chemical Society Published on Web 03/21/1998

3020 J. Am. Chem. Soc., Vol. 120, No. 13, 1998 Scheme 1. Synthesis of the Modified DNA Dimersa

a Dimers shown were subsequently converted to phosphoramidates and incorporated into DNA oligomers.

imine. The difference in the thermal melting temperature was as high as 15 °C,5c but the structural and energetic origins of the altered stability have not been defined.2-5 In contrast with the explosive growth in synthesizing backbonemodified oligonucleotides, only limited progress has been made in determining the structural and energetic factors which control the stability of the duplexes.3 De Mesmaeker et al.2e,f have reviewed the chemical modifications and the stability of the second generation antisense oligonucleotides that contain no phosphorus atom in the modified linkages. The hybridization stability of these duplexes, as measured by the thermal melting temperature, is rationalized in terms of the change in charge, conformational rigidity, structural preorganization, hydrophobicity, and steric hindrance differences (see Figures 5 and 6 and Table 1 in ref 2f). In principle, structural studies by NMR and X-ray can provide important insight into the origins of the changes in enthalpy and entropy resulting from backbone modifications. However, the thermodynamic analyses depend heavily on the thermal melting temperature,2d-f which involves a complex interplay between the enthalpic and entropic components of the duplexes and/or the single strands.3 We saw the following issues relating to the structure and stability of the backbone-modified oligonucleotides as critical to any such analysis: (1) Since the formation of a nucleic acid duplex is highly cooperative, could a chemical modification in the middle of one strand change the coopertivity of duplex formation, deviating, for example, from the usual two-state thermal transition? (2) Can the thermodynamic stability be dissected into changes in enthalpy and entropy of the modified duplex relative to its native analogue? (3) Is the difference between the stability of the modified duplex and the native duplex caused

Luo et al. by a change in the duplex structure, the modified single strand, or both? Although the comparisons between a backbonemodified duplex with its native analogue have generally assumed that only the duplex stability, rather than the single strand, has been changed upon backbone modification, the experimental basis for such an implicit assumption is not yet clear. (4) What is the overall structure in the modified duplex as compared to its native analogue? Is it still nativelike, retaining the general features characteristic of the native duplex, or is the structure significantly altered? If the overall structure of the native duplex is preserved upon backbone modification, is it possible to detect structural changes around the modified region? (5) Finally, can any global and local structural changes be related to changes in ∆S and ∆H values? As a specific example to illustrate the points listed above, we have studied the short DNA duplex, 1, formed between d(CpGpTNTpGpC) and its complementary d(GpCpApApCpG) strand. A single phosphodiester is replaced by the aminoethyl group,5 and the approach taken is similar in spirit to studies of point mutations on protein stability. A comparison is made of the melting curves obtained for duplex 1, and its analogues, with the native DNA sequence. Both UV and NMR analyses, under differing salt and sample concentration conditions, are employed, and the thermal unfolding curves are dissected into changes in enthalpy and entropy. For the native DNA duplex, quantitative comparisons are made among values derived from UV and NMR measurements as well as from calculated values from the literature. NMR measurements show that the structure of duplex 1 is nativelike and maintains the Watson-Crick base pairs as exhibited by similar cross-peak patterns characteristic of the native duplex. However, significant perturbation in the modified TT region is detected using the NOE cross-peaks between the two methylenes of the aminoethyl linkage and the flanking sugar protons. This result allows for a rationalization of the changes in enthalpy and entropy of the backbone-modified duplex to the observed local and global structural changes. These results have allowed a model to be proposed which attempts to explain the observed destabilization of backbone-modified oligonucleotides with neutral linkers. One important implication underlying this model, as well as others proposed recently,6 is the role that the polyanionic phosphate backbone plays in conferring duplex stability. The comprehensive approach that we have taken should be useful for analyzing other backbonemodified nucleic acids and allow the factors that control duplex stability to be better defined. Results Synthesis of Modified Nucleosides. The hexameric template, d(GpCpApApCpG), was designed to be small, palindromic, and non-self-complementary, and to have a high G/C content to enhance association.3 The central template adenosine residues simplified the synthetic chemistry to modifications of thymidine (Scheme 1). The ease of radical allylation at C-3′ (60% recrystallized yield over three steps) set the stereochemical position of the aldehyde that was ultimately generated by oxidative cleavage.3a Only the generation of the 5′-amine was required on the second thymidine derivative. For bulk systhesis, reductive amination of the 3′-carboxyaldehyde 5 with the 5′amine 7 gave the amine-linked dimer 8 in >95% purified yield.3a,7 Acetylation of 8 gave cleanly the N-acyl dimer 9, and further oxidation of 5 to 6 and condensation with 7 gave the amide dimer 10. Each of these dimers was converted into the (6) Richert, C.; Roughton, A. L.; Benner, S. A. J. Am. Chem. Soc. 1996, 118. 4518.

Stability of the DNA Duplex with Backbone Modifications

J. Am. Chem. Soc., Vol. 120, No. 13, 1998 3021

Scheme 2. Modified DNA Duplex with the TT Dimer Region As Indicateda

Figure 1. ln(CT/4) vs 1/Tm for the native duplex in 500 mM NaCl, 10 mM NaH2PO4, 2 mM NaN3, pH 6 buffer.

to eq 2a,8a,b or alternatively from a nonlinear least-squares fit to eq 2b, where Cd is the duplex concentration. The nonlinear

a Three linkages in the TT dimer are shown: 1, aminoethyl; 2, N-acetylaminoethyl; 3, amide.

appropriate phosphoramidite for incorporation into the hexanucleotide via solid-phase synthesis. This approach permitted the ready synthesis of 1 as well as the modified hexamers 2 and 3 shown in Scheme 2. Thermodynamic Measurements. Strand association was analyzed with both UV and NMR spectroscopies, and different methods for extracting the thermodynamic values were explored. In all cases, attempts were made to fit the data to a simple equilibrium between the DNA duplex and the single strands. In this two-state model, the equilibrium constant, K, for association of non-self-complementary strands is expressed by eq 1, where CT is the total concentration of oligonucleotide strands and f is the fraction of strands in the duplex form.8a,b

K ) 2f/[(1 - f)2CT]

(1)

UV Analyses. The melting temperature, Tm, of the native duplex was determined via the hypochromic shift of the base transition at five different concentrations in 500 mM NaCl. The ∆H and ∆S values were derived from the slope and intercept, respectively, of a plot of ln(CT/4) vs 1/Tm in Figure 1, according (7) Caulfield, T. J.; Prasad, C. V. C.; Prouty, C. P.; Saha, A. K.; Sardaro, M. P.; Schairer, W. C.; Yawman, A.; Upson, D. A., & Kruse, L. I. Bioorg. Med. Chem. Lett. 1993, 3, 2771. (8) (a) Borer, P. N.; Dengler, B.; Tinoco Jr, I.; Uhlenbeck, O. C. J. Mol. Biol. 1974, 86, 843. (b) Albergo, D. D., Marky, L. A., Breslauer, K. J.; Turner, D. H. Biochemistry 1981, 20, 1409. (c) Albergo, D. D.; Turner, D. H. Biochemistry 1981, 20, 1413.

ln(CT/4) ) (∆H/R)(1/Tm) - (∆S/R)

(2a)

Cd ) 2 exp((∆H/R)(1/Tm) - (∆S/R))

(2b)

least-squares analysis has the advantage of fitting the thermodynamic functions and baseline parameters simultaneously to the primary data. This approach avoids the empirical assumption of the linear baseline regions at the low- and high-temperature limits. Both approaches gave identical values of ∆H and ∆S (Table 1); however, as expected, the errors are much larger with the least-squares fit. At a 67% confidence level, the error was 7 times larger. A similar increase in error was previously noted by Santoro and Bolen9 in determinations of the ∆G of protein unfolding. NMR Analyses. The design of the template placed two adenine bases in the center of the duplex. The two aromatic methyls of the thymidine bases in the complementary strand, T3-CH3 and T4-CH3, reside in the center of the duplex and flank the site of the structural change that is under consideration. The chemical shifts of these methyl protons in the native and the modified duplex 1 show cooperative thermal transitions with temperature only when measured in the presence of the template strand (Figure 2a). In addition to being able to monitor the center of the duplex, these NMR analyses have several other advantages. First, the higher concentrations of the nucleic acids used in the NMR experiments move the Tm for these short strands above room temperature in H2O. Second, the baselines for the duplex, δd, and the single strands, δs, are well defined (Figure 2a). Finally, the exchange rate between the two states is fast on the NMR time scale. When the observed chemical shift, δobs, is modeled as a function of these two states, eq 3,10 the resulting van’t Hoff plots from eq 4, Figure 2b, give ∆H

δobs ) (1 - f)δs+ fδd

(3)

ln K ) -(∆H/R)(1/T) + (∆S/R)

(4)

and ∆S values for both the T3-CH3 and T4-CH3 groups that (9) Santoro, M. M.; Bolen, D. W. Biochemistry 1988, 27, 8063. (10) (a) Tran-Dinh, S.; Neumann, J.-M.; Huynh-Dinh, T.; Genissel, B.; Igolen, J.; Simonnot, G. Eur. J. Biochem. 1982, 124, 415. (b) Pieter, J.; Mellema, J.-R.; Van Den Elst, H.; Van Der Marel, G.; Van Boom, J.; Altona, C. Biopolymers 1989, 28, 741.

3022 J. Am. Chem. Soc., Vol. 120, No. 13, 1998

Luo et al.

Table 1. Thermodynamic Parameters of Thermal Unfolding from the Native Duplex and the Backbone-Modified Duplex 1 to Coiled Strandsa thermodynamic parameters

predictedb (native)

UV meltc (native)

T3CH3d (native)

44.4

43.6 ((1.8)

47.3 ((1.3)

-∆H (kcal mol-1) difference in -∆H between native and modified duplex 1 -∆S (eu mol-1) difference in -∆S between native and modified duplex 1

126.1

T3CH3d (duplex 1)

T4CH3d (native)

T4CH3d (duplex 1)

40.2 ((0.9) 46.1 ((2.5) 39.1 ((1.7) 7.1 7.0 120.2 ((5.8) 136.9 ((4.0) 121.4 ((3.0) 131.9 ((8.2) 117.1 ((5.7) 15.5 14.8

a The results are derived from the van’t Hoff plots in Figures 1b and 2b. b Predicted values were calculated using the nearest-neighbor thermodynamic data11a at 1 M NaCl, pH 7. The 126.1 eu mol-1 was derived from the predicted ∆H and ∆G using the nearest-neighbor approximation. c UV melting at different DNA concentrations in 500 mM NaCl, 10 mM NaH2PO4, 2 mM NaN3, pH 6. d The change in chemical shift with temperature as monitored by T3 and T4 methyl groups at 2 mM total oligonucleotide concentration in 100 mM NaCl, 10 mM NaH2PO4, 2 mM NaN3, pH 6.

Figure 2. (a, top) Chemical shift change of the T3 methyl groups in the native (b) and the modified duplex 1 (O) as a function of temperature. The curves through data points are fitted to eq 5 using the nonlinear least-squares method. Lines through each melting profile are fit with either floating or fixed baseline parameters in δs and δd. The difference in Tm between native and modified duplex 1 is ca. 17 °C in 2 mM solutions in 9:1 (v/v) H2O-D2O, 100 mM NaCl, 10 mM NaH2PO4, 2 mM NaN3, pH 6 buffer. (b, bottom) van’t Hoff plots of the thermal unfolding profiles for the native ([) and modified duplex 1 (]).

are in close agreement with those determined from UV analysis (Table 1). By combining eqs 1 and 3, the change in chemical shift can be fit to eq 5, where T is the temperature in Kelvin, R

δobs ) δs + (δd - δs)(1 + c - ((1 + c)2 - 1)1/2) c ) 1/(CTK) ) (1/CT) exp(∆H/(RT) - ∆S/R)

(5)

is the gas constant, and δs, δd, ∆H, and ∆S are the parameters in the nonlinear least-squares analysis. By this method, the T3CH3 gives ∆H ) -45.5 kcal mol-1 and ∆S ) -131 eu mol-1, in general agreement with the direct van’t Hoff analysis, but again the error is 8-9 times greater. Holding the baseline parameters of δd and δs at the values used in the van’t Hoff analysis, this error is reduced 3-fold, consistent with the floating baselines being largely responsible for the errors.9

Justifying the Analyses. The calculated ∆H value for this sequence, using the nearest-neighbor approximation of Breslauer et al.,11a is -44.4 kcal mol-1. These estimated ∆H values for short oligonucleotide duplexes give excellent agreement with calorimetry measurements.8a,b,11a The ratio of the UV-determined ∆H to this calculated value is 0.98, justifying the use of the two-state assumption for duplex melting.8b Although UV analyses have been used extensively for measuring the global unfolding of DNA duplexes, the ∆δ of the thymidine methyl resonances may be diagnostic of a more localized perturbation. These methyls, which reside in the center of the duplex, yield ∆H values that are of the same order, -47.3 kcal mol-1 from T3-CH3 and -46.1 kcal mol-1 from T4-CH3, but slightly more negative than those from calculations (Table 1). The more negative values may result from the cooperative transition monitored only at the center of the duplex, free from the fraying of base pairs at the ends, or from the isotope effect in D2O.8b The ∆S value was also calculated from the predicted ∆G and ∆H of Breslauer et al.11a to be -126 eu mol-1 in 1 M NaCl. By UV in 500 mM NaCl, ∆S ) -120 ( 5.8 eu mol-1; by NMR in 100 mM NaCl, ∆S ) -137 ( 4.0 and -132 ( 8.2 eu mol-1 from T3-CH3 and T4-CH3, respectively. These differences in ∆S are accounted for quantitatively by the difference in salt concentrations used in the experiments. According to the polyelectrolyte theory of nucleic acids,12 the salt effect is primarily of entropic origin that arises from counterion condensation on the phosphate backbones of the duplex, and this assignment is supported by precise scanning microcalorimetry measurements.13 By substituting the concentration of NaCl into eq 6, where C is the molar salt concentration and R is the gas constant, the change in entropy due to the salt effect on duplex

∆S ) (-1.11 + 0.36 log C)R eu mol-1 of phosphate

(6)

melting is predicted as 2.9, 2.4, and 2.2 eu mol-1 of phosphate at 100 mM, 500 mM, and 1 M NaCl, respectively. If the NaCl concentration is changed from 100 mM to 1 M, the increase in ∆S upon duplex formation is predicted to be 0.7 eu mol-1 of phosphate or 7 eu mol-1 of a hexamer duplex containing 10 phosphate groups. The predicted increase of 7 eu mol-1 in ∆S upon duplex formation going from 100 mM to 1 M NaCl lies within the observed difference range of 11 eu mol-1 from T3CH3 and 5.8 eu mol-1 from T4-CH3 by NMR and the calculated ∆S at 1 M NaCl (Table 1). The values measured by UV differ slightly from the predicted values. This general agreement among the predicted and UV- and NMR-determined ∆H and (11) (a) Breslauer, K. J.; Frank, R.; Blo¨cker, H.; Marky, L. A. Proc. Natl. Acad. Sci. U.S.A. 1986, 83, 3746. (b) Vesnaver, G.; Breslauer, K. J. Proc. Natl. Acad. Sci. U.S.A. 1991, 88, 3569. (12) (a) Manning, G. S. Q. ReV. Biophys. 1978, 11, 179. (b) Record, Jr. M. T.; Mazur, S. J.; Melancon, P.; Roe, J.-H.; Shaner, S. L.; Unger, L. Annu. ReV. Biochem. 1981, 50, 997. (13) Privalov, P. L.; Filimonov, V. V. J. Mol. Biol. 1978, 122, 447.

Stability of the DNA Duplex with Backbone Modifications

J. Am. Chem. Soc., Vol. 120, No. 13, 1998 3023

∆S values for the native duplex suggests that the errors associated with these measurements are too small to account for the large difference in the enthalpy and entropy between the native and the modified duplex 1 (see the discussion below). The Modified Duplex. By UV analyses, the modified duplex 1 shows a cooperative melting transition at both 10 and 100 µM; however, the melting temperature is reduced ∼17 °C relative to the native duplex. At 2 mM, the NMR-detected thermal transitions of both duplex 1 and the native duplex are more cooperative, and the Tm of duplex 1 is 17 °C lower than that of the native duplex (Figure 2a). Each thymidine methyl reported the thermal transition, and more significantly, the difference in the thermodynamic values determined from T3CH3 and T4-CH3 for the native and the modified duplex 1 were the same (Table 1). By this analysis, the overall decrease in stability of the modified duplex is seen to originate from two opposing factors: (i) a large decrease in enthalpy, 7 kcal mol-1, and (ii) a favorable change in entropy of 15 eu mol-1. The positive charge in the backbone is expected to contribute enthalpic stabilization to duplex 1 due to the more favorable Coulombic interaction along the oligonucleotide backbone.14a The entropic contribution resulting from the counterion condensation on the modified backbone is expected to be unfavorable relative to its native analogue. Therefore, the observed changes in both ∆S and ∆H values between native and modified duplex 1 are in the opposite direction from those expected by electrostatic considerations. The N-acyl- and amide-linked hexamers of duplexes 2 and 3 were prepared as shown in Scheme 1 to remove the positive charge in the backbone. At a 5 µM duplex concentration in 1 M NaCl, the neutral backbone of the N-acyl duplex 2 showed no cooperative thermal melting transition. The N-acyl group in duplex 2 alters both the steric bulk of the linkage and the hybridization of the N atom; however, duplex stability was tolerant to substituents on the nitrogen atom at the corresponding position of an amide linkage when the group was changed from H to methyl to isopropyl.2f,14b Given the increased conformational flexibility of the ethylamine linkage of duplex 1 in comparison with the corresponding amide linkage in 3 (amide analogues 14a to 14c in Table 1 of ref 2f), the acyl substituent is expected to be well tolerated in 2. Amide duplex 3 was prepared also, and its Tm of 14 °C, only slightly less than the Tm (18 °C) of the native DNA duplex, was consistent with the observation made by De Mesmaeker et al.2f,14b These data suggest that the positive charge is not the major factor contributing to duplex stability, and other factors should be sought to explain the overall destabilization of duplex 1 relative to the native structure. Structural Analyses. Figure 3 shows the resonances of the exchangeable imino protons in 9:1 H2O-D2O for both the native DNA and the modified duplex 1. All six of the expected protons are present in both structures, consistent with their participation in hydrogen bonding with the complementary strand. As the temperature is increased, the protons on the modified strand begin to exchange with solvent approximately 20 °C lower than do those in the native DNA. In both cases, fraying of the ends is followed by an apparent cooperative disruption of the internal base pairs. These protons are assigned using the standard approach that follows the sequential connectivities involving exchangeable imino, amino, and adenine-H2 protons by 1D NOE difference spectra.15 The 1D spectra made it possible to obtain a sufficient

signal/noise ratio by accumulating many scans at the specific resonances (see the Materials and Methods). Figure 4 shows that irradiation of the imino proton of T4 results in two sharp H2 NOEs; a strong H2 NOE with its base-pairing partner A3 and a weaker H2 NOE with A4. Only one strong H2 NOE with A4 is observed when the imino proton of T3 is irradiated (Figure 4). These are characteristic of the Watson-Crick basepairing motif in the B-form DNA duplex for both the native duplex (Figure 4a) and the modified duplex 1 (Figure 4b).15a To the extent that the Watson-Crick base-pairing motif is

(14) (a) Dempcy, R. O.; Browne, K. A.; Bruice, T. C. J. Am. Chem. Soc. 1995, 117, 6140. (b) De Mesmaeker, A.; Waldner, A.; Lebreton, J.; Hoffmann, P.; Fritsch, V.; Wolf, R.; Freier, S. Angew. Chem., Int. Ed. Engl. 1994, 33, 226.

(15) (a) Chou, S. H.; Hare, D. R.; Wemmer, D. E.; Reid, B. R. Biochemistry 1983, 22, 3037. (b) Chou, S. H.; Flynn, P.; Reid, B. R. Biochemistry 1989, 28, 2435. (c) Weiss, M. A.; Patel, D. J.; Sauer, R. T.; Karplus, M. Nucleic Acids Res. 1984, 12, 4035.

Figure 3. Temperature dependence of exchangeable imino proton resonances for (a, top) the native and (b, bottom). the modified duplex 1. The exchange-out temperature of the hydrogen-bonded imino protons is higher by ca. 20 °C in the native duplex relative to the modified duplex 1. The same samples as in Figure 2 are used, and the results are consistent with a ca. 17 °C difference in Tm between both duplexes shown in Figure 2a.

3024 J. Am. Chem. Soc., Vol. 120, No. 13, 1998

Luo et al.

Figure 4. 1D NOE difference spectra showing the imino and AH2 proton connectivities for (a, top) the native and (b, bottom) the modified 1 DNA duplexes in the T3T4/A4A3 base pairs. Irradiation of the imino proton assigned to T4 results in two sharp H2 NOEs: a strong H2 NOE with its pairing partner A3 and a weaker H2 NOE with A4. Only one strong H2 NOE with A4 is observed when the imino proton of T3 is irradiated, consistent with a right-handed duplex. The NOE connectivities for imino and adenine H2 protons in the T3T4/A4A3 base pairs indicate that the Watson-Crick base-pairing motif is preserved in the backbone-modified DNA duplex 1.

preserved in the native duplex, it is similarly maintained in the backbone-modified duplex 1. Characterization of the Modified Duplex Conformation via Nonexchangeable Protons. Figure 5a shows the characteristic NOESY cross-peaks for the sequence-specific connectivities between nonexchangeable base H8/H6 and sugar H2′/ 2′′ protons in the modified DNA duplex 1.16 The solid lines show the sequence-specific assignments of the backbonemodified TNT strand. Each base proton H8/H6 displays NOEs with its own and 5′-flanking sugar H2′/2′′ protons and/or 3′flanking methyl groups involving thymidine. Starting from the 5′-terminal nucleotide with only two NOEs of the 5′-base proton with its own sugar H2′/2′′ protons, the sequential connectivities from base H8/H6 to sugar H2′/2′′ protons were obtained across the entire strand without disruption. The dashed lines in Figure 5a show the corresponding assignments for the complementary native DNA strand in the modified duplex 1. The intranucleotide NOEs of H1′ and H3′ with H2′/2′′ are utilized to distinguish the H2′ from H2′′ protons on the basis of the variations in NOE cross-peak intensities. The NOE cross(16) (a) Scheek, R. M.; Boelens, R.; Russo, N.; van Boom, J. H.; Kaptein, R. Biochemistry 1984, 23, 1371. (b) Wu¨thrich, K. NMR of Proteins and Nucleic Acids; Wiley & Sons: New York, 1984.

Figure 5. Expanded regions of the NOESY spectrum (D2O, pH 6, 1 °C, 200-ms mixing time) showing (a, top) the sequence-specific connectivities between nonexchangeable base H8/H6 and sugar H2′/ 2′′ protons for the modified DNA duplex 1. The sequential connectivities for the d(CpGpTNTpGpC) or TT strand and its complementary d(GpCpApApCpG) or AA strand are shown in solid and dashed lines, respectively, with the cross-peaks designated by the base and its sequence number. (b, bottom) NOE connectivities of the intranucleotide sugars H1′ and H3′ with their H2′/2′′ protons in the modified DNA duplex 1. The reversal in the chemical shift of H2′/2′′ protons in T3 are revealed by a stronger cross-peak of the high-field signal with the T3-H1′ proton. The T3-H3′ proton is shifted upfield (see Figure 6) due to the chemical modification in the T3 sugar 3′ position.

peak intensity of H1′ with H2′′ is stronger than with H2′ because the H1′-H2′′ distance is usually shorter under all sugar pucker conformations. For the same reason, the corresponding NOE cross-peak intensity of the H3′ with H2′′ is weaker than that of H3′ with H2′. On the basis of these criteria, the chemical shifts of H2′ and H2′′ protons in the 3′-modified T3 sugar were assigned unambiguously even though their resonance frequencies are reversed in comparison with the usual order of H2′ and H2′′ (Figure 5), except for the H2′ and H2′′ protons of 3′terminal G6 with a free 3′-OH group (Figure 5b). These assignments are important in defining the orientation of the methylene groups in the modified backbone with respect to the

Stability of the DNA Duplex with Backbone Modifications

J. Am. Chem. Soc., Vol. 120, No. 13, 1998 3025

Table 2. Proton Chemical Shift Assignments for the Modified DNA Duplex 1 at pH 6, 1 °Ca residue

H6/H8

G1 C2 A3 A4 C5 G6

8.01 7.51 8.30 8.28 7.30 7.93

C1 G2 T3 T4 G5 C6

7.73 8.09 7.46 7.53 8.04 7.49

a

H5/CH3 5.43 5.29 5.92 1.49 1.72 5.98

H1′

H2′

H2′′

H3′

H4′

H5′/H5′′

5.99 5.48 5.96 6.12 5.59 6.17

2.63 2.13 2.81 2.65 1.89 2.62

2.79 2.40 2.97 2.84 2.30 2.36

4.87 4.90 5.10 5.07 4.80 4.68

4.27 4.18 4.35 4.42 4.16 4.21

3.73/3.73 4.12/4.10 4.15 4.23 4.26 4.08

5.79 6.11 5.86 5.97 5.99 6.22

2.15 2.78 2.27 1.72 2.65 2.22

2.51 2.91 1.86 2.37 2.77 2.22

4.75 5.01 2.78 4.58 5.07 4.53

4.11 4.26 4.05 4.35 4.45 4.27

3.76 4.14/4.06 4.42/4.27 3.71/3.56

other

AH2 7.43 AH2 7.74

C6′ 2.02; C7′ 3.39

4.08

Referenced to TSP at 0.00 ppm.

Figure 6. Expanded region of the NOESY spectrum (D2O, pH 6, 1 °C, 200-ms mixing time) which highlights the cross-peaks of C7′ and C6′ methylene protons on the modified linkage with flanking sugar protons. Notice the stronger C7′-T3-H1′ cross-peak relative to those of C6′-T3-H1′ and even T3-H1′/H2′′. The cross-peaks are assigned as the sugar protons and methylene groups are in Scheme 1.

T3 sugar ring. Figure 5b also indicates the chemical shift overlap for H2′ and H2′′ in C6 as shown by a single strong cross-peak between H1′ and H3′ with H2′/2′′, respectively. The base-pairing scheme and sequential connectivities, with the uniformly much stronger NOEs of an intraresidue base proton with its own H2′/2′′ protons than the corresponding NOEs with the 5′-flanking interresidue H2′/2′′ protons across both strands, establish a canonical right-handed B-DNA conformation for the backbone-modified DNA duplex 1.16 The chemical shift assignments of the modified DNA duplex 1 are listed in Table 2. NOEs of the Modified Linkage with Flanking Sugar Protons. The orientation of the phosphodiester linkage relative to the flanking nucleosides has eluded direct NOE observation in the DNA duplex due to the lack of protons and severe overlap of the 5′,5′′ resonances.17 The presence of the methylene units designated as C6′ and C7′ in the modified linkage of 1 permitted detailed analysis of the backbone geometry using the corresponding NOEs of these protons with the 5′-flanking sugar protons. These data are shown in a well-resolved region of the NOESY spectrum in Figure 6 and are summarized in Table 3 for the NOE cross-peaks of the C6′ and C7′ protons with the T3-H1′, -H2′′, and -H4′ sugar protons underneath the T3 sugar (17) (a) Gao, X.; Brown, F. K.; Jeffs, P.; Bischofberger, N.; Lin, K.-Y.; Pipe, A. J.; Noble, S. A. Biochemistry 1992, 31, 6228. (b) Gao, X.; Jeffs, P. W. J. Biomol. NMR 1994, 4, 17.

ring. In contrast to the corresponding PO2- group that lies on the outside of the native duplex with maximal solvent accessibility, the C7′ CH2 group in the modified duplex 1 can maintain close contact with the hydrophobic duplex interior and the T3 sugar ring. This orientation is supported by the stronger NOE cross-peak of T3-H1′-C7′ (H1′-C7′; ca. 2.7 Å) as compared to the T3-H1′-C6′ cross-peak (H1′-C6′; ca. 3.2 Å) and the NOE cross-peak of C7′ with T3-H2′′ in Figure 6. In comparison, the corresponding distances of the O1P and O2P atoms of a PO2- group to the T3-H1′ proton, or H1′(i)-O1P(i + 1) and H1′(i)-O2P(i + 1), are 5.2-5.5 and 4.8-5.4 Å, respectively, in a canonical B-DNA conformation,17,18 exceeding the detectable distance limit of the NOE (ca. 4.5-5.0 Å). An NOE cross-peak between the H1′ proton and a C7′ methylene has been observed in one other backbone modification (see Figure 9 in ref 17a), although it was suggested that the magnetization exchange may have occurred by spin diffusion in this case. The T3-H1′-C7′ cross-peak is indicative of an altered backbone orientation in duplex 1 as the presence of this crosspeak was confirmed in an NOESY spectrum collected at a 50ms mixing time (see the Supporting Information), making spin diffusion unlikely as a major contributor to the observed crosspeaks.19 Table 3 tabulates proton-proton distances derived from NOE cross-peaks that are characteristic of the backbone conformation in the modified region. For comparison, the phosphate oxygenproton distances in a native B-form DNA, but with a T3 sugar pucker in either C2′-endo or C3′-endo forms, are also listed. The difference between the distances of T3-H1′ to C7′ methylene of 1 and T3-H1′ to OP* of the native duplex is too large to be adequately accounted for by changes in the connecting sugar puckers (see Table 3). These data argue that the modified backbone of duplex 1 deviates significantly from the backbone conformation of the canonical DNA ensembles. Structural Model. A preliminary structural model was derived by computer simulations with ca. 180 constraints derived from NMR experiments. A semiquantitative approach was taken to generate a 3D model using the Biosym software package.20 The initial structure was built and modified from a B-DNA duplex using the Biopolymer and Builder modules in InsightII.21 The energy minimization and dynamic simulations were carried out using the Discover module in InsightII with the CVFF force (18) (a) Chandrasekaran, R.; Birdasall, D. L.; Leslie, A. G. W.; Ratcliff, R. L. Nature 1980, 283, 743. (b) Dickerson, R. E. J. Mol. Biol. 1983, 166, 419. (19) Reid, B. R.; Banks, K.; Flynn, P.; Nerdal, W. Biochemistry 1989, 28, 10001. (20) Dwyer, T. J.; Geierstanger, B. H.; Mrksich, M.; Dervan, P. B.; Wemmer, D. E. J. Am. Chem. Soc. 1993, 115, 9900. (21) Arnott, S.; Hukins, D. W. L. J. Mol. Biol. 1973, 81, 93.

3026 J. Am. Chem. Soc., Vol. 120, No. 13, 1998

Luo et al.

Table 3. Comparisons of the Proton-Proton Distance Derived from NOE Cross-Peaks Characteristic of the Backbone Conformation in the Modified Region with the Corresponding Phosphate Oxygen-Proton Distances in B-Form DNA but with the T3 Sugar Pucker in Either the C2′-endo or C3′-endo Conformation i to i + 1a

H1′-OP* b

H2′-OP *

H2′′-OP*

H3′-OP*

H4′-OP*

distances in B-DNAc distances in twisted B-DNAd T3 sugar to C7′ methylene NOE derived distancese NOE cross-peaks at 50 ms

4.8-5.5 4.7 H1′-C7′ 2.7 medium

3.7-5.3 3.8 H2′-C7′ ∼4.5 null

2.9-4.8 2.5 H2′′-C7′ 3.2 weak

2.4-3.8 3.9 H3′-C7′ >3.5 very weak

4.7 2.8 H4′-C7′ 2.7 medium

a i and i + 1 refer to two consecutive residues. b OP* refers to the pseudo atom for two O1P and O2P atoms. c Distance ranges for B-DNA were taken from Table 4 in ref 17a. d The twisted B-DNA represents the extreme case of the T3 sugar pucker switching from normal C2′-endo to C3′-endo. e The NOE-derived distances were obtained using the integration of the NOE cross-peak at 50-ms and 200-ms mixing times; the NOE cross-peak intensities at 50-ms mixing time provided the control on the distance estimates.

Figure 7. Comparison of structures with their solvent-accessible surfaces (dots) in the TT region of (a) the ethylamine-containing strand, and (b) its native counterpart in B-form DNA. Both strands were cut out of the full duplexes, generated from crystal structure data for (b) and from molecular dynamics simulations incorporating NMR constraints for (a). The intruding C7′ methylene with the flat solvent-accessible surface in the modified linkage 1 contrasts markedly with the maximally solvent-accessible phosphate of the phosphodiester linkage.

field parameters. The hydrogen bonds in the Watson-Crick base pairs were constrained as supported by the imino proton experiments to prevent the two strands from breaking apart during simulations. The simulations were started by using the simulated annealing protocol to satisfy local NOE constraints in the TT dimer region while the rest of the structure was constrained. Once the NOE constraints in the TT dimer region were satisfied, the whole structure was refined to remove steric clashes with all experimental constraints using dynamic calculations and energy minimization to generate the preliminary 3D model. It should be pointed out that DNA is an extended, nonglobular molecule, and NMR can only provide short-range NOE distance constraints (